Thomas Kollar

Head of AI @ Wayve; ex-Research Manager @ TRI; ex-Amazon; ex Apple. CMU / MIT

[Wayve @wayve_ai]

Meet LA-Pose. Our latest model taking Wayve another step towards generalization at scale.

LA-Pose employs large-scale self-supervised learning, building strong motion representations for 3D perception from 10.2 million unlabeled driving video snippets, unlike today's strongest approaches that often depend on expensive, carefully curated 3D supervision.

With only a lightweight pose head and limited labelled data, LA-Pose achieves:

📷 State-of-the-art camera pose estimation
🌎 Strong zero-shot generalization across diverse driving scenarios
🏷️ Orders of magnitude less labelled data than fully supervised 3D approaches

Our full blog post: https://t.co/CcNWuLHJsn
Explore the full paper here: https://t.co/DHRsAS9ckV

[Nissan Motor @NissanMotor]

Nissan and #Wayve have signed a partnership agreement that will bring our next-gen #ProPILOT driver assistance tech powered by Wayve #AI to a broad range of #Nissan vehicles.

Nissan aims to first launch the next-gen tech in Japan in fiscal year 2027.
https://t.co/FbFB2LU8VI

[Wayve @wayve_ai]

GAIA 3 introduces four powerful new capabilities that unlock richer and more scalable evaluation of autonomous driving systems. 🌍

🧵 Follow the thread below to see examples of;

1. Long perturb generations 🚗
2. Safety augmentations ⚠️
3. Semantic augmentations 🌤️🌅🌙
4. Embodiment transfer 🚘📷

GAIA 3 re-generates the same scenario as if observed from different vehicles with different camera positions. One scene, three embodiments, consistent dynamics. Ideal for testing models across different hardware setups.

These advances show how GAIA-3 brings new realism, diversity, and scale to the evaluation of end-to-end driving systems. 🚀

Dive into the full blog: https://t.co/pIk8xG1ENe

Every clip you see below is generated by GAIA-3.👇

#GAIA3 #EmbodiedAI #AISafety #GenerativeAI #AutonomousVehicles

[Jamie Shotton @Jamie_Shotton]

Big things cooking in Tahoe... 🚀

[Thomas Kollar @tkollar]

@siddkaramcheti @ICatGT @GTrobotics @mlatgt Congrats @siddkaramcheti on joining Georgia Tech! 🚀 Seeing your growth through our collaborations and the time at @ToyotaResearch was truly amazing. I can’t wait to see what you build next at @GTrobotics @ICatGT @mlatgt . Well deserved! 👏

[Jamie Shotton @Jamie_Shotton]

It's awesome to be back in the Bay Area this week at @wayve_ai's other North American office.
I can't wait to test the massive progress the team's been making on rides around the Bay Area and city while I'm here, and to meet with our science leaders @vijaycivs @tkollar @gianlucacorrado and others to galvanise the groups at the start of an incredibly exciting #YearOfEmbodiedAI ahead!

#Science #Team #EmbodiedAI

[Thomas Kollar @tkollar]

@achalddave With @sedrickkeh2 @achalddave @karora4u @MercatJean @vslevic @sy_gadre

[Thomas Kollar @tkollar]

Building language models is difficult and requires high quality preprocessing, modeling, evaluation and large scale training.

As significant collaborators in this project at TRI, the resulting 7B model DCLM-7B is a significant achievement. It is a competitor to Mistral 7B and LLaMA-7B, even though trained on less data. And it’s fully open. And that’s just the start of the competition.

Excited to see how others leverage these results to build even more capable language models and improve dataset quality.

[Vaishaal Shankar @Vaishaal]

I am really excited to introduce DataComp for Language Models (DCLM), our new testbed for controlled dataset experiments aimed at improving language models. 1/x

[Thomas Kollar @tkollar]

More details from @achalddave:

https://t.co/JOEFBXgrqj

[Achal Dave @achalddave]

Check out DataComp for language models! Open data, open code, open training recipe, and close to Llama3-8B performance. This has been a labor of love over the last year, a huge thanks to all the collaborators for helping make this happen!

[Thomas Kollar @tkollar]

https://t.co/EYwuJitYRk More info on Prismatic here.

[Siddharth Karamcheti @siddkaramcheti]

What design choices matter when developing a visually-conditioned language model (VLM)?

Check out our paper – Prismatic VLMs – and open-source training code, evaluation suite, and 42 pretrained VLMs at the 7B-13B scale!

📜 https://t.co/yyarNk7GuZ
⚙️ + 🤗 https://t.co/TsoQGsuSN2

[Thomas Kollar @tkollar]

Excited to release Prismatic!

Cutting through the noise of vision-language modeling, Prismatic is a release of 42 pre-trained VLMs from the 7B to 13B scale, a codebase for rigorous evaluation and a myriad of insights for what matters for performance.

[Siddharth Karamcheti @siddkaramcheti]

What design choices matter when developing a visually-conditioned language model (VLM)?

Check out our paper – Prismatic VLMs – and open-source training code, evaluation suite, and 42 pretrained VLMs at the 7B-13B scale!

📜 https://t.co/yyarNk7GuZ
⚙️ + 🤗 https://t.co/TsoQGsuSN2

[Thomas Kollar @tkollar]

By first developing some of the best Vision-Language Models with Prismatic at TRI:
https://t.co/Se8oDRVSBp

OpenVLA was able to quickly build some of the best generalist policies for robotics. Code, data and weights are all open-source:
https://t.co/Lp6DlvvTpr

This is a great achievement! Congrats @moo_jin_kim @siddkaramcheti @KarlPertsch @ashwinb96 @SurajNair_1 and all collaborators.

[Moo Jin Kim @moo_jin_kim]

✨ Introducing 𝐎𝐩𝐞𝐧𝐕𝐋𝐀 — an open-source vision-language-action model for robotics! 👐

- SOTA generalist policy
- 7B params
- outperforms Octo, RT-2-X on zero-shot evals 🦾
- trained on 970k episodes from OpenX dataset 🤖
- fully open: model/code/data all online 🤗

🧵👇

[Thomas Kollar @tkollar]

@roydanroy 😂

[Sedrick Keh @sedrickkeh2]

Recurrent models like RWKV and Mamba have gained attention recently, but these can be costly to train and iterate on.

What if we could simply...
turn Mistral/Llama/Gemma into an RNN? 🎩🪄

Presenting our work, Linearizing Large Language Models!

https://t.co/hbaSUWk8uc

[Thomas Kollar @tkollar]

@MercatJean @sedrickkeh2 @achalddave @vslevic @karora4u @adnothing @sy_gadre Additional collaborators include @archit_sharma97 @lschmidt3 @ericmitchellai and many more.

[Thomas Kollar @tkollar]

Over the last year at TRI we’ve been training Large Language Models, including results in the following areas:
Scaling: https://t.co/zkbTfwpGkz
Alignment: https://t.co/zchTwZCndy

As a part of upcoming work, we are sharing back with the open source community and releasing a performant Mamba model that we’ve trained at the 7B parameter scale. More results on linear transformers upcoming.

[Sedrick Keh @sedrickkeh2]

📢 Releasing TRI's open-source Mamba-7B trained on 1.2T tokens of RefinedWeb!

Mamba-7B is the largest fully recurrent Mamba model trained and is a state-of-the-art recurrent LLM. 🚀🚀🚀

https://t.co/PmsoRc4SNG

[Thomas Kollar @tkollar]

With @MercatJean @sedrickkeh2 @achalddave @vslevic @karora4u @adnothing @sy_gadre for the release.

[Thomas Kollar @tkollar]

@roydanroy @yoavgo This makes me want to read the paper.